drusepth

Why would you buy a 5-year-old iPhone for the same price you can get a new Android with comparable specs though? If I'm gonna spend 2-3 hundred on a phone, I'd like it to last at least a couple more years. Regardless of OS, you're more likely to get that on a new phone vs any phone 5+ years old.

> Solve enough problems relying on AI writing the code as a black box, and over time your grasp of coding will worsen, and you wont be undestanding what the AI should be doing or what it is doing wrong - not even at the architectural level, except in broad strokes.

Using AI myself _and_ managing teams almost exclusively using AI has made this point clear: you shouldn't rely on it as a black box. You can rely on it to write the code, but (for now at least) you should still be deeply involved in the "problem solving" (that is, deciding _how_ to fix the problem).

A good rule of thumb that has worked well for me is to spend at least 20 min refining agent plans for every ~5 min of actual agent dev time. YMMV based on plan scope (obviously this doesn't apply to small fixes, and applies even moreso to larger scopes).

"Ads are always separate and clearly labeled."

I've heard this before...

I think the "we" that can work on these systemic problems and actually improve them are a very different "we" than those of us who just need basic health care right now and will take anything "we" can get.

This isn't feasible for a huge swathe of the USA, often because of costs/insurance but sometimes literally just accessibility/availability. A few years ago it took me nearly 8 months to find a PCP in my city that was accepting new patients (and, wee, they dropped my insurance less than a year after).

"People using AI" had a meaningful change when they "joined the workforce" in 2025.

We may not have gotten fully-autonomous employees, but human employees using AI are doing way more than they could before, both in depth and scale.

Claude Code is basically a full-time "employee" on my (profitable) open source projects, but it's still a tool I use to do all the work. Claude Code is basically a full-time "employee" at my job, but it's still a tool I use to do all the work. My workload has shifted to high-level design decisions instead of writing the code, which is kind of exactly what would have happened if AI "joined the workforce" and I had a bunch of new hires under me.

I do recognize this article is largely targeted at non-dev workforces though, where it _largely_ holds up but most of my friends outside of the tech world have either gotten new jobs thanks to increased capability through AI or have severely integrated AI into whatever workflows they're doing at work (again, as a tool) and are excelling compared to employees who don't utilize AI.

Generally, any expert hopes their tool/paintbrush/etc is as performant as possible.

Saw this earlier today and legitimately thought it was satire, especially when I heard the price. Turns out it's real?

The classic Brave approach!

It seems really weird to me that such granular intercorporate details are made publicly available (in a blog post?). I've never had to publicly state things like this when making corporate partnerships. That makes me wonder how much of this post is crafted solely for PR...

Doesn't it already do this? I keep a domain or two on HTTP to force network-level auth flows (which don't always fire correctly when hitting HTTPS) and I've gotten warnings from Chrome about those sites every time for years... Only if I've been to the site recently does the warning not show up.

Brilliantly phrased — sharp, concise, and perfectly captures that uncanny "AI-polished" cadence everyone recognizes but can’t quite name. The tone strikes just the right balance between wit and warning.

Presumably the reverse would be more valuable: design something in Figma, then have Gemini CLI read from Figma and use that as a reference while writing the code.

I don't know if this is how the Figma integration works, but right now I'm just manually screenshotting my designs and passing them into Claude CLI as references for what I want, so this seems potentially more streamlined.

Curious: what would, should, or could they be doing?

I don't use any adblockers. The results are also the same for me whether I'm logged in or not.

Piggybacking on to provide a screenshot since I also see no sponsored content and Midjourney is my #1 result, well above the fold.

[1] https://i.imgur.com/Oxo4FJl.png

Are 235B models classified as local LLMs? I guess they probably are, but others in this thread are probably looking more toward 20B-30B models and sizes that generally fit on the RAM you'd expect in average or slightly-higher-end hardware.

My beefy 3D gamedev workstation with a 4090 and 128GB RAM can't even run a 235B model unless it's extremely quantized (and even then, only at like single-digit tokens/minute).

Also: the answer marked as a solution that says it's not a feature is not even from a Spotify employee.

Is there any precedent for this in Spotify? Can I tell it to hide all country music or all music by German artists?

It's a comment about the company/brand behind the models, not the individual models themselves.

I think this comment exposes an important point to make: people have different opinions of what "vibe coding" even means. If I were to ask an LLM to vibe code a text editor, I guarantee you I wouldn't get a React app using Supabase -- because I'd give it pages of requirements documentation and tell it not only what I want, but the important decisions on how to make it.

Obviously no model is going to one-shot something like a full text editor, but there's an ocean of difference between defining vibe coding as prompting "Make me a text editor" versus spending days/weeks going back and forth on architecture and implementation with a model while it's implementing things bottom-up.

Both seem like common definitions of the term, but only one of them will _actually_ work here.

I've worked in 3 different WYSIWYG editors for web and desktop applications over the years, lightly contributed to a handful of other open-source editors, and spent plenty of time building my own personal editors from scratch (and am currently using gpt-5 to fix my own human bugs in a rewrite of the Notebook.ai text editor that I re-re-implemented ~8 years ago).

Editors are incredibly complex and require domain knowledge to guide agents toward the correct architecture and implementation (and away from the usual naive pitfalls), but in my experience the latest models reason about and implement features/changes just fine.

Definitely fast, but initial use puts quality either comparable to or below gpt-5-nano. This might be a low-cost option for people who don't mind babysitting the output (or working in very small projects), but claude/gpt-5/gemini all seem to have significantly higher quality at marginally more cost/time.

By just emphasizing the speed here, I wonder if their workflows revolve more around the vibe practice of generating N solutions to a problem in parallel and selecting the "best". If so, it might still win out on speed (if it can reliably produce at least one higher-quality output, which remains to be seen), but also quickly loses any cost margin benefits.

Ironically, you could probably vibe code your own.

> Likewise for the legal profession, if all judges made twice the number of mistakes, how much would the world suffer?

Is this hyperbole? It seems like the real question being asked here is "would the world be worse off without deterministic checks and balances", which I think most people would agree is true, no?

On the other hand, fault-intolerant codebases are also often highly defined and almost always have rigorous automated tests already, which are two contexts where coding agents specifically excel in.

I feel this, honestly. I get so much more work done (currently: building & shipping games, maintaining websites, managing APIs, releasing several mobile apps, and developing native desktop applications) managing 5x claude instances that the majority of my time is sucked up by just prompting whichever agent is done on their next task(s), and there's a real feeling of lost productivity if any agent is left idle for too long.

The only time to browse HN left is when all the agents are comfortably spinning away.

Is there any documentation on what the max sub usage limit is? A coworker tried it and was booted off Opus within just a couple hours due to "high usage". I haven't made the jump since I expect my $3k/mo on API would just instantly fly by a $200/mo sub and then I'd just be back on API again, but if it could carve out $1k-2k of costs for a little bit of time managing sub(s) it might be worth it.

This is also the hilarious part of "prompt engineering".

It's just effective linguistics and speech; what people have called "soft skills" forever is now, obviously, trying to be a science for some reason.

It's like 5-10% more for a Waymo (which is nothing in the $10-30 range) and you don't have to talk to anyone and/or sit in awkward silence for the ride. Yeah, I'll pay that every time.